Test Data Analysis:

Large Model: awpainting_v12.jpg (base training model)

Lora Weights: Range from 0.2 to 0.8 is acceptable

Sampler: Euler/DPM++ 2M SDE/DPM++ 3M (Optimal: Euler)

Large Model: toonyou_beta6.safetensors

Lora Weights: Range from 0.2 to 0.6 is acceptable

Sampler: Euler/DPM++ 2M SDE (Optimal: Euler, with Lora 0.8 also performing well)

The Lora model training was conducted with rigorous selection of approximately 190 illustrations focused on landscapes, outdoors, oceans, and lakes as the theme for the training materials. Leveraging the Kohya training method, the model has successfully calculated a novel illustration style in a unique and highly accurate manner. In practical applications, it can rapidly and precisely generate landscape illustrations based on specific needs, significantly enhancing work efficiency and improving the quality of creations. For designers, this undoubtedly provides a brand-new style of painting and creative expression.